Advancements in Safe Legged Locomotion: IRIM Symposium

Advancements in Safe Legged Locomotion: IRIM Symposium

Table of Contents:

1. Introduction
2. Background
3. Research Focus
4. Temporal Logic-Based Navigation
5. Safe Locomotion in Partially Observable Environments
6. Hierarchy and Heterogeneous Teaming of Robots
7. Navigation in Human-Crowded Environments
8. Multi-Contact Locomotion and Manipulation
9. Contact Sensing and Soft Arm Integration
10. Angular Momentum-Aware Trajectory Optimization
11. Task and Motion Planning for Push Recovery
12. Conclusion

Introduction

In this article, we will explore the recent works of Professor Zhao, an assistant professor in mechanical engineering. He will be discussing the advancements his research group has made in the field of robotic navigation and coordination. The focus will be on how robots can safely navigate and coordinate in challenging and dynamically changing environments. Professor Zhao's group has a particular interest in developing computationally efficient trajectory optimization and formal methods to handle contact behaviors and ensure safety, robustness, and collaborative behaviors.

Background

Before diving into the specific research projects, let's understand the background of Professor Zhao's work. He heads the Laboratory for Intelligent Decision and Autonomous Robots, which focuses on task and motion planning, decision making, and control algorithms for highly agile robotic systems. The group's research is centered around locomotion and manipulation, with a special emphasis on challenging problems that involve contact behaviors and formal guarantees of safety and robustness.

Research Focus

The main area of focus for Professor Zhao's research group is to develop robust and scalable task and motion planning algorithms. The goal is to enable safe robot locomotion in complex and human-populated environments. The group aims to design trajectory optimization algorithms that can handle uncertainties caused by contact behaviors and generalize these optimizations to versatile contact behaviors using a universal regularization-based policy learning approach. The researchers are particularly interested in navigation in partially observable environments, which requires reactive and proactive interactions with both obstacles and humans.

Temporal Logic-Based Navigation

One of the key research areas explored by Professor Zhao's group is temporal logic-based formal methods for navigation. They have been studying this approach from multiple perspectives, including whole-body locomotion planning, safe learning planning, and heterogeneous robot teaming. By using temporal logic-based formal methods, the group can specify complex navigation tasks involving manipulations and collisions with dynamic obstacles. This approach allows for reactive synthesis and ensures compliance with task specifications.

Safe Locomotion in Partially Observable Environments

An important research question tackled by Professor Zhao's group is how to guarantee safe robot navigation in partially known environments while accomplishing complicated tasks. They address the challenge of balancing safety and navigation by designing safe locomotion gates based on reduced-order models of the robots. These gates incorporate safety criteria that determine the allowable changes in the robot's APEX velocity and turning angle. By encoding these criteria into the high-level task planning, the group ensures that the robot's decisions Align with the underlying motion planners and controllers.

Hierarchy and Heterogeneous Teaming of Robots

To enhance the capabilities of robot navigation and coordination, Professor Zhao's group proposes a hierarchical and heterogeneous teaming approach. By combining bipedal and aerial robots, they tackle environmental conflicts that arise at runtime. For example, when a drone encounters a door blocking its path, it can request assistance from a walking robot to open the door. This online replanning framework enables the detection of failure cases and the synthesis of motion behaviors to address conflicts dynamically.

Navigation in Human-Crowded Environments

The group is also exploring the challenges of navigation in human-crowded environments. To ensure safe interactions with humans, they aim to infer human navigation intentions and predict their navigation paths using state-of-the-art neural network approaches. The researchers plan to incorporate this human predictor module into their model predictive control (MPC)-based robot motion planner, enhancing the robot's risk sensitivity and enabling proactive interactions with humans.

Multi-Contact Locomotion and Manipulation

Another area of interest for Professor Zhao's group is multi-contact locomotion and manipulation. They focus on solving optimal task allocation and planning problems for object-centric contact planning. The researchers adopt a signal temporal logic approach to specify collaborative locomotion and manipulation task specifications. These specifications are then transformed into mixed-integer programs for multi-contact modeling. By treating contact uncertainties as disturbances, the group uses reactive synthesis to ensure the robot's compliance with the specifications.

Contact Sensing and Soft Arm Integration

In their latest project, Professor Zhao's group is studying contact sensing and reach with a bipedal robot equipped with a soft super-limb arm. The robotic fingers are equipped with high-resolution gel sets tactile sensing, enabling both locomotion and manipulation capabilities. The researchers are exploring how to integrate rich contact sensing information into planning and decision-making processes to enhance the robot's capabilities.

Angular Momentum-Aware Trajectory Optimization

Trajectory optimization is a significant focus of Professor Zhao's group. They have developed an angular momentum-aware trajectory optimization algorithm that achieves dynamic Consensus of centroidal momentum and whole-body optimization. By considering the ellipsoidal moment of inertia of the robots and gathering feedback from the whole-body dynamics, the algorithm enables the robot to perform a range of dynamic motions, including 180-degree spins.

Task and Motion Planning for Push Recovery

The group has also developed a framework for task and motion planning for push recovery that is resilient to various perturbations. The framework ensures that the robot can recover from perturbations of any direction and magnitude during different locomotion phases, such as walking, turning, or running. They have evaluated the framework extensively in realistic scenarios, including human-augmented spaces, to assess its effectiveness.

Conclusion

In conclusion, Professor Zhao and his research group are making significant advancements in the field of robotic navigation and coordination. Their focus on safe locomotion, hierarchical planning, multi-contact manipulation, and human interactions showcases the breadth of their work. By considering challenging environments, incorporating formal methods, and integrating advanced sensing capabilities, the group is pushing the boundaries of humanoid robotics.

Highlights:

• Professor Zhao's research group focuses on safe navigation and coordination for robots in challenging environments.
• They develop computationally efficient trajectory optimization algorithms and formal methods to handle contact behaviors and ensure safety and robustness.
• The group is particularly interested in navigation in partially observable environments and proactive interactions with humans.
• Their research involves temporal logic-based navigation, safe locomotion gates, hierarchical planning, and heterogeneous robot teaming.
• They explore navigation in human-crowded environments, multi-contact locomotion, sensor integration, and trajectory optimization.
• Professor Zhao's group aims to transform humanoid robots from a science fiction concept to a reality.

FAQ:

Q: What is the main focus of Professor Zhao's research group? A: Professor Zhao's research group focuses on developing robust and scalable task and motion planning algorithms for safe robot navigation and coordination in challenging environments.

Q: How does their research address navigation in partially observable environments? A: They incorporate reactive and proactive interactions with obstacles and humans to ensure safe navigation in partially observable environments.

Q: What is the significance of hierarchical and heterogeneous teaming of robots? A: This approach allows for dynamic replanning and conflict resolution in environments where conflicts arise at runtime.

Q: How do they ensure safe locomotion in their navigation framework? A: They design safe locomotion gates based on reduced-order models to balance safety and navigation requirements.

Q: How does Professor Zhao's group integrate contact sensing into planning and decision-making processes? A: They explore the integration of rich contact sensing information from gel sets and soft arm technologies to enhance robot capabilities.

Q: What is the focus of their trajectory optimization research? A: They develop angular momentum-aware trajectory optimization algorithms and incorporate dynamic consensus of centroidal momentum and whole-body optimization.

Q: How does their framework for push recovery address perturbations? A: Their framework is resilient to perturbations of any direction and magnitude during different locomotion phases, ensuring efficient recovery.

Q: What are the future directions of their research? A: They are exploring navigation in human-crowded environments, incorporating human predictor modules, and developing multi-contact locomotion and manipulation techniques.